April 14 2017 ERCOT Astrapé Consulting LLC The Brattle Group 1 Introduction Workshop Agenda 3 Workshop Goals Provide stakeholders with a solid understanding of the EORMMERM modeling approaches ID: 815533
Download The PPT/PDF document "Economically Optimum Reserve Margin (EOR..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Economically Optimum Reserve Margin (EORM) WorkshopApril 14, 2017ERCOT, Astrapé Consulting LLC, The Brattle Group
Slide21. Introduction
Slide3Workshop Agenda3
Slide4Workshop GoalsProvide stakeholders with a solid understanding of the EORM/MERM modeling approaches
Consensus decision-making on methods for EORM/MERM studies
Stay
focused on methodological approaches and issues; we will not discuss opportunities for
supplementary
studies
4
Discuss each method used for 2014 Brattle/Astrapé study and NERC Loss of Load studies
Collaboratively modify the method
Identify unsatisfied concerns
Assess the degree of support
Finalize the method, or table for further discussion at follow-up conference call
Slide5Summary of 2014 Optimal Reserve Margin StudyIn 2014 the PUCT asked Brattle and Astrapé to
estimate the economically-optimal reserve margin in ERCOT to inform their ongoing review of market design for resource adequacy (full study
linked here
)
Under base case assumptions, we
estimated
reserve margins*
of:
10.2% economic optimum
11.5% in energy-only equilibrium (also minimizes customer cost)
14.1% required to meet 1-in-10 Enforcing a 1-in-10 reserve margin requirement
at 14.1% (with or without a capacity market) would increase long-run average customer costs by approximately $400 million/year or 1% of retail rates relative to the current energy-only market in equilibrium (considering energy and capacity price impacts, not considering other costs and benefits)
A 14.1% reserve margin requirement would also reach equilibrium price levels sooner than an energy-only market, meaning near-term impacts would be greater
5
*2014 reported reserve margins used ELCC and other reserve margin accounting conventions at the time. Updated conventions would increase all reported numbers by 1.8%.
Slide6Summary of 2014 Optimal Reserve Margin Study
We estimated a 14.1% reserve margin required to meet the traditional 1-in-10 loss of load event (LOLE) standard, results sensitive to:
Forward period
at which supply decisions are locked in, and consequential load forecast error (LFE) that needs to be considered in analysis (removing LFE drops the reserve margin to 12.6%)
Likelihood of 2011 weather
recurring, with this extreme year treated at 1% chance in base case (increasing to 1/15 or equal chance would increase the reserve margin to 16.1
%)
Estimated other reliability metrics as well
6
14.1% for 1-in-10
9.6% for 0.001% Normalized EUE
Reliability Metrics Across Reserve Margins
*
*2014 reported reserve margins used ELCC and other reserve margin accounting conventions at the time.
Slide72. Modeling the Electricity System
Slide8SERVM Modeling Framework
Hourly chronological unit commitment and dispatch
Stochastic (Monte Carlo)
simulation involving thousands of model iterations
Simulate distributions for load/weather, load growth uncertainty, outages, fuel prices, intermittent renewable output, demand response, etc.
Simulate Emergency Operating Procedures
Simulate scarcity pricing during shortage events
Example based on recent Loss of Load studies conducted for ERCOT:
Weather
(13 years of weather history)
Economic load forecast error
(distribution of 5 points)Multi-state unit outage modeling, capturing frequency and duration (50 iterations)Probabilistic modeling of external assistance/DC ties, Private Use Network generating units
Total load scenario breakdown: 13 weather years x 5 LFE points = 65 scenarios; Total iteration breakdown: 65 scenarios * 50 unit outage iterations = 3,250 yearly simulations or until convergence is metModel run-time:
A single 50 iteration scenario takes ~ 30 minutes. These scenarios are simulated in parallel across multiple cores8
Slide9Supply
Slide10Thermal ResourcesModeled based on CDR and ERCOT internal production cost model data
Seasonal Capacity (Vary by hourly temperature)
Heat Rate Curves
VOM
Startup Costs
Fuel Prices
Startup Times
Emissions
Min-up/Min-down Times
Ramp Rates
Ancillary Service Capability AGC CapableQuick Start
10
Slide11Forced/Planned Thermal Generator OutagesFull Outages
Time to Repair
Time to Failure
Partial Outages
Time to Repair
Time to Failure
Derate
Percentage
Startup Failures
Maintenance Outages
Planned Outages
Created based on Outage Scheduler data for December 2010 – July 201611
Unit Name
Capacity Weighted Equivalent Forced Outage Rate (%)
Nuclear
4.04%
Coal
6.32%
Gas Combined Cycle
4.27%
Gas Combustion Turbine
19.42%
Gas Steam Turbine
20.03%
Fleet
Capacity Weighted
Average
EFOR
8.50
%
*Values based on ERCOT’s latest Loss of Load study for NERC (
http://www.nerc.com/pa/RAPA/ra/Reliability%20Assessments%20DL/2016ProbA_Report_Final_March.pdf
)
Slide12Forced/Planned Thermal Generator Outages12
Slide13Private Use Network Resources13
Draw Probability
10%
10%
10%
10%
10%
10%
10%
10%
10%
10%
Load Level
Net Output (MW)
Above 100% of Peak Forecast
3,631
3,800
3,822
3,875
3,885
3,885
3,916
3,927
3,991
4,090
95% - 100% of Peak Forecast
3,042
3,222
3,403
3,583
3,764
3,809
4,026
4,243
4,459
4,676
90% - 95% of Peak Forecast
3,042
3,201
3,360
3,519
3,677
3,723
3,938
4,154
4,370
4,585
85% - 90% of Peak Forecast
2,633
2,860
3,087
3,314
3,541
3,587
3,825
4,063
4,302 4,540 70% - 85% of Peak Forecast 2,270 2,531 2,792 3,053 3,314 3,360 3,655 3,950 4,245 4,540 Below 70% of Peak Forecast 999 1,044 1,067 1,441 2,497 2,542 2,985 3,428 3,870 4,313
Net Output modeled stochastically
Most
Recent NERC Study: Output based on
load level as shown below
Previous EORM study: Output based on market price
Slide14Hydro Resources13 years
of
historical hydro energy modeled
Peak
shaving capacity based
on regression above
14
Slide15Hydro Resources15
Peak Shaving Capability
Maximum Output during August
Annual Energy
521 MW nameplate
Characterized resources based on:
4 years of hourly data from ERCOT
15 years of monthly data from FERC 923
Hydro resources modeled with different parameters each month:
Monthly total energy output
Daily peak shaving capability (from historical hourly data, recreated for other years based on
regression analysis)
Energy output is partly peak shaving (scheduled during daily load peaks) and partly run-of-river (output in off-peak hours)
Remaining hydro capacity (nameplate minus output) counts toward RRS and ORDC x-axis (approximately the same as 240 MW of
hydro-synchronous
resource that ERCOT assumes in PRC calculation)
Slide16Switchable, Mothballed, Retired Units
Past studies based
on
CDR forecast/accounting rules
Switchable Units
:
Included as internal
resources with capacity deductions based on Notices of “Unavailable Capacity” submitted by SGR owners
Propose adopting a price response method similar to how PUN units are modeled
Retirements
:
Exclude starting in the CDR-specified year based on Notices of “Suspension of Operations of a Generation ResourceSeasonal Mothballs:
included 1,876 MW of seasonal mothballs that are available for dispatch only in summer months, nearly all of them from May-September Permanent Mothballs: Exclude
from modelPlanned New Units: Excluded from model to achieve low starting reserve margin in simulations; then increase with marginal unit
16
Slide17Wind ModelingHourly shape by weather yearDeveloped by third party consultant by site
Aggregated to CDR values for study year
17
Slide18Solar ModelingHourly shape by weather yearDeveloped by third party consultant by site
Aggregated to CDR values for study year
18
Slide19Demand-Side ResourcesDispatched based on price subject to the following
Hours per season
Time of day
Weekday/Weekend
19
Slide20Demand-Side Resources20
Summer Capacity
Call Limits
Call Priority
TDSP Standard Load Management Programs
208
16 hours per year, during hours 14-20
1
Load Resources Serving as Responsive Reserve
1,153
Unlimited
2
10 Min ERS
609
8 hours per season and per hourly intervals;
Seasons: Winter, Spring, Summer, Fall;
Hourly intervals: week day hours 1-8 and 21-24 and weekends, week day hours 9-13, week day hours 14-16, week day hours 17-20
3
30 Min ERS
898
8 hours per season and per hourly intervals;
Seasons: Winter, Spring, Summer, Fall;
Hourly intervals: week day hours 1-8 and 21-24 and weekends, week day hours 9-13, week day hours 14-16, week day hours 17-20
4
Slide21Demand
Slide22Weather Uncertainty on Loads22
Peak Load by Weather Year
(Before and After DR Gross-Up)
Load Duration Curves
(Peak Hours, Before DR Gross-Up)
Slide23Economic Growth Uncertainty
Non-weather load forecast error (LFE) increases with forward period
Assume normally distributed forecast error with standard deviation of 0.8% on a 1-year forward basis, increasing by 0.6% with each additional forward year
Scale of error is a standard assumption developed in lieu of an ERCOT-specific analysis
We assume no bias or asymmetry in the non-weather LFE (unlike weather-driven LFE which has greater upside than downside uncertainty)
Modeling approach:
Assume resource decisions and reserve margin must be “locked in” 3 years forward, so realized forecast error is larger than if more shorter-term supply options were available
Sensitivity analysis examining impact of forward periods ranging from 0 to 4 years forward
23
Non-Weather Forecast Error
With Increasing Forward Period
3-Year Forward LFE
Discrete LFE Error Points Modeled
Slide24Transmission
Slide25Transmission Topology and Import/Export ModelingExternal Regions
Hourly Loads
Resources
DC
Ties (probabilistic for EORM study; based on historic flows during highest peak load hours for NERC Loss of Load study)
Share resources based on
economics
25
810 MW
280 MW
Mexico
(Coahuila, Nuevo Leon, & Tamaulipas)
ERCOT
5,180 MW
Entergy
(MISO)
SPP
Slide26Generation Resource Mix
26
Values based on previous EORM study
Slide27Diversity with External Regions27
Values based on previous EORM study
Slide28The Market
Slide29Representation of Energy and Ancillary Service MarketsEconomic reserve margin analyses need to simulate market dispatch and prices
Most important elements to model are those that characterize scarcity event:
System cost
(drives the optimal reserve margin)
Energy/ancillary prices
(drives the market equilibrium reserve margin
)
Recommend using a similar approach in future studies, but reviewing the need for significant market design updates
29
Unit Commitment
Week-ahead unit commitment
4-hour ahead unit commitment
Hourly dispatch of 10 & 30-min quickstart
Energy MarketHourly production cost model representing real-time energy market
Prices at marginal cost, considering generation and demand responseDetailed representation of scarcity event costs and pricing (see next slides)Scarcity prices affected by ancillary service shortage pricing
Ancillary Services
Regulation, spinning, and non-spinning reservesOperating Reserve Demand Curve (ORDC)
Power Balance Penalty Curve (PBPC)
Slide30Scarcity Pricing and Emergency Procedures
30
Emergency Level
Marginal
Resource
Trigger
Price
Marginal
System Cost
n/a
Generation
Price
Approximately $20-$250
Same
n/a
Imports
Price
Approximately $20-$250
Up to $1,000
during load shed
Same
n/a
Non-Spin Shortage
Price
Marginal Energy +
Non-Spin ORDC
w/ X = 2,000
Marginal Energy +
Non-Spin ORDC w/
X = 1,150
n/a
Emergency Generation
Price
$500
Same
n/a
Price-Responsive Demand
Price
$250-$9000
Same
n/a
Spin Shortage
Price
Marginal Energy + Non-Spin
+
Spin ORDC
w/ X = 2,000
Marginal Energy + Non-Spin +
Spin ORDC w/
X = 1,150
n/a
Regulation Shortage
Price
Power Balance Penalty Curve
Same
EEA 1
30-Minute ERSSpin ORDC x-axis = 2,300 MW$3,239 at Summer Peak(from ORDC)$1,405EEA 1
TDSP Load Curtailments
Spin
ORDC x-axis
=
1,750 MW
$9,000 (from ORDC)
$2,450
EEA 2
Load
Resources in RRS
Spin
ORDC x-axis
=
1,700 MW
$9,000 (from ORDC)
$2,569
EEA 2
10-Minute
ERS
Spin
ORDC x-axis
=
1,300 MW
$9,000 (from ORDC)
$3,681
EEA 3
Load Shed
Spin
ORDC x-axis
=
1,150 MW
VOLL = $9,000
Same
Recommend reviewing 2014 study assumptions for material updates.
Slide31Operating Reserve Demand Curve31
ORDC is one of the most important drivers of economic reserve margin
Important to distinguish the distinction between:
Marginal system cost
(blue curve) based on ERCOT’s analysis of the likelihood of lost load when running short of reserves
Price-setting ORDC
(red curve) at X = 2,000 used to set market prices
Recommend implementing updated ORDC curves (4 seasons, 6 times of day, 2 reserve types) and updating for design enhancements as needed
Operating Reserve Demand Curves
Example: Summer Hours 15-18
Sources and Notes:
Page 35, Figure 17, Brattle EORM
report.
Slide32Selection of the Marginal Resource
Basic study approach is to add (or subtract) supply from the model until we find the cost-minimizing quantity (optimal reserve margin) or the maximum quantity that the market will attract (market equilibrium), see later slides
Outcome depends partly on what type of resource is added
If the market is near equilibrium among resource types, results will be similar regardless of this choice (e.g., 2014 study found similar results whether using gas CC or CT as the marginal resource)
Recommend future studies:
Use gas CC as the default type, given the greater quantity and uniformity of CC technologies
Maintain flexibility to examine CTs as a test, or examining other types if system conditions change
32
2014 Reserve Margin Study Results
Gas CC Marginal
Gas CT Marginal
Slide33Cost of New Entry Estimate33
CONE
is
a significant driver of results, with higher CONE driving lower economic reserve margin
Subject to significant uncertainties, e.g., ATWACC
2014 study adapted PJM’s bottom-up engineering CONE estimate, applying adjustments for ATWACC, location, and interconnection costs
Recommend developing a forum for Market Participants to request CONE updates or sensitivity analyses; update process would consider formal CONE estimates from other markets and EIA, then use judgement to apply reasonable adjustments
Performance Characteristics
Sources and Notes:
Performance consistent with 2011 Brattle ERCOT Study after update of merchant ATWACC assumption.
One year of escalation adapted from 2013 ISO-NE ORTP parameter estimates.
2011 & 2014 Gross CONE Values
Slide34Value of Lost Load (VOLL)Value of lost load (VOLL) is the cost imposed on customers when they face involuntary service interruptions
VOLL is subject to significant uncertainty depending on customer type and study approach. Estimates can range $1,000 to $100,000/MWh* (but high-end numbers not appropriate to use in this study since these high-value end uses should already have back-up power to protect against distribution outages)
In our 2014 study, we found that a VOLL range of $4,500 to $18,000/MWh translated to a 8.9%-11.8% range in the economically optimal reserve margin (that study also scaled DR curtailment costs in proportion to VOLL)
Recommend maintaining 2014 study approach that used the VOLL assumed in ERCOT nodal protocols and used as the High System-Wide Offer Cap, or $9,000/MWh
34
Notes
:
*Approximately $1,500
– $3,000/MWh for residential, $10,000 – $50,000/MWh for commercial, and $10,000 – $80,000/MWh for industrial loads according to a
MISO survey. See MISO “Value of Lost Load Final Report. May 15, 2006.
3. Study Results
Slide36Reserve Margin Accounting
CDR method:
Firm Load = Peak Load Forecast grossed up for Energy Efficiency
– Load Resources (ERS, providing Responsive Reserves)
– Energy Efficiency
Total Resources =
Thermal (Seasonal Net Sustained Capability)
+ Hydro peak capacity contribution
+
Wind/Solar seasonal peak average contribution
+ Switchable Capacity less amount unavailable
+ Available mothballed and RMR capacity
+ PUN capacity contribution forecast
+ DC Tie capacity contribution + Planned thermal resources + Planned wind/solar seasonal peak average contribution
36
Slide37Reliability Based Reserve Margin TargetsNERC Assessment Metrics
LOLH
EUE
EUE / Net Energy for Load
Industry Standard
LOLE
1 day in 10 year standardEquivalent to 0.1 LOLE
37
Slide38Reliability Based Reserve Margin Targets38
Reserve Margins Required to Meet Alternative Physical Reliability Standards
Base Case
Slide39Reliability Based Reserve Margin TargetsEORM Study
39
Base Case
Assumptions:
1
% chance of 2011 weather
Load forecast error consistent with 3 years forward
CC as marginal technology
14.1%
reserve margin for 0.1 LOLE
Sensitivities
If 2011 is given equal weather weight (1/15 chance), the reserve margin increases to
16.1%
If non-weather LFE is excluded, the reserve margin drops to
12.6% LOLE w/ Differing 2011 Weather Weight
LOLE w/ Varying Forward Periods
Slide40Renewable Effective Load Carrying Capability
ELCC Simulations: Not conducted for NERC Assessment or previous EORM
Average ELCC of wind and solar
Calibrate system to
0.1 LOLE
Remove entire wind/solar fleet and determine conventional capacity needed to maintain 0.1 LOLE
Average ELCC = conventional capacity added/ nameplate capacity of wind/solar fleet
Incremental ELCC of
wind (coastal or non coastal)
and
solar (fixed vs. tracking)Start at 0.1 LOLE
Add 1,000 MW of wind and remove conventional MW to maintain 0.1 LOLEELCC = Conventional MW removed / 1,000 MW of wind40
Slide41Renewable Capacity Contributions
Use historical output rather than Effective Load Carrying Capability (ELCC) method
Current
CDR
capacity contribution methodology for
wind
resources
Develop a “Seasonal Peak Average” capacity factor (WINDPEAK%) by season and coastal/non-coastal regions
Based
on average historical availability (defined as a generator’s telemetered High Sustained Limit, or HSL) during the highest 20 seasonal peak load hours for an historical period of up to 10
yearsCapacity
contribution values are re-calculated after each season with new seasonal historical dataCoastal and non-coastal resources are calculated separately due to the significantly different diurnal wind patterns for these regionsMultiply WINDPEAK% by installed nameplate capacityFor
solar resources, similar approach; calculations done system-wide with only preceding three years of historical seasonal data
41
Slide424. STUDY PROCESS AND SCOPE
Slide43Study Design Elements
EORM/MERM study updated every two years starting in mid-2018; aligns with current project schedule for conducting NERC’s biennial Probabilistic Assessments
Single simulation year, four years beyond study year (e.g., 2018 study would simulate the year 2022)
Continue to use CDR load and capacity variable definitions outlined in Protocols sections 3.2.6.2.1/ 3.2.6.2.2
Include all CDR planned resources with projected CODs as of December 31 of the year prior to the study year
Although not part of the EORM/MERM study effort, NERC requires a “Reference Margin Level” that serves as a resource adequacy performance benchmark
If ERCOT declines to provide NERC with a Reference Margin Level, NERC will use a 15% default value
43
Slide44Economically Optimal Reserve Margin Results44
Economically Optimal
Reserve Margin at 10.2%
Total System Costs across Planning Reserve Margins
Notes:
Total
system costs include a large baseline of total system costs that do not change across reserve margins, including $15.2 B/year in transmission and distribution, $9.6 B/year in fixed costs for generators other than the marginal unit, and $10B/year in production costs.
Slide45Market Equilibrium Reserve Margin Results45
Slide46Purpose and Scope of Potential Sensitivity Analysis and ScenariosDefinitions
Sensitivity analysis:
U
sed to test robustness and uncertainty in a particular finding given uncertainties in individual parameters
Scenario analysis:
Used to test the performance of decisions (e.g. transmission plans) against fundamentally different futures
For the purposes of the 2018 EORM/MERM study, ERCOT and stakeholders should focus only on
sensitivity analysis
for key drivers
Develop a short prioritized list for consideration
46
Potential Sensitivities
(Drivers of EORM & MERM)
Forward period for load forecast uncertaintyWeighting of extreme weather years Gross Cost of New Entry
Marginal resource typeGas pricesValue of Lost Load (VOLL)
Quantity of intermittent resources
Slide47Vetting of Study Results
ERCOT Proposal
Develop a study plan and project schedule to be presented to the Supply Analysis Working Group (SAWG)
Study plan should document any proposed deviations from the study methodology document and explain the rationale
Present the draft EORM/MERM study report to the SAWG, Wholesale Market Subcommittee, and PUCT staff for review and comment
Establish a new document section on the Resource Adequacy Webpage at ercot.com for making the final study report, associated public data, and report comments available for download
Codify, to the extent needed, this vetting process in the Nodal Protocols
47
Slide48Periodic Methodology Assessments and Updates
48
ERCOT Proposal
During the study cycle off-year (2019, 2021, etc.), hold one or more SAWG meetings, with PUCT staff participation, by the end of Q3 to discuss the need for methodology updates
If any updates are warranted, ERCOT will modify the EORM/MERM methodology document by year end
ERCOT and SAWG meeting participants may agree on the need for sensitivity analysis prior to approving a change
Changes may include:
CDR-related Protocol revisions
Market-design-related Protocol revisions
Changes in key cost parameters such as CONE and VOLL
Source and derivation of model inputs (not the inputs themselves)
Updated methodology document posted to the Resource Adequacy Webpage at least two months prior to the start of the next study
Slide495. Next Steps
Slide50Next Steps and Schedule
Agree on the need for a follow-up WebEx conference call (approximately mid-May or as
determined by
workshop participants)
Request for written follow-up comments: send them to Pete Warnken,
Pete.Warnken@ercot.com
Establish a sensitivity analysis work plan based on workshop participant requests, analysis scope/complexity, and
prioritization
; use the SERVM model and NERC 2016 LOL study data set
Provide PUCT staff and SAWG participants with periodic updates on EORM/MERM methodology document development progress
Complete and post the document by the end of 2017, contingent on the sensitivity work plan
50